Epidemiology – measuring

disease frequency
Nikolai Hristov, MD, PhD
The science of epidemiology
• Epidemiology as defined by Last is “the study of the distribution and
determinants of health-related states or events in specified populations, and
the application of this study to the prevention and control of health problems”.
• Comparing rates of disease in subgroups of the human population became
common practice in the late nineteenth and early twentieth centuries. This
approach was initially applied to the control of communicable diseases, but
proved to be a useful way of linking environmental conditions or agents to
specific diseases.
• In the second half of the twentieth century, these methods were applied to
chronic non-communicable diseases such as heart disease and cancer,
especially in middle and high-income countries.
Measuring health and disease
• The measure of health and disease is fundamental to the practice of
epidemiology.
• A variety of measures are used to characterize the overall health of
populations.
• The most ambitious definition of health is that proposed by WHO in 1948:
“health is a state of complete physical, mental, and social well-being and not
merely the absence of disease or infirmity.” This definition – criticized because
of the difficulty in defining and measuring well-being – remains an ideal.
• Practical definitions of health and disease are needed in epidemiology, which
concentrates on aspects of health that are easily measurable and amenable to
improvement.
Definitions
• Definitions of health states used by epidemiologists tend to be simple,
for example, “disease present” or “disease absent”.
• For example, guidelines about cut-off points for treating high blood
pressure are arbitrary, as there is a continuous increase in risk of
cardiovascular disease at every level.
• Diagnostic criteria are usually based on symptoms, signs, history and
test results. For example, hepatitis can be identified by the presence
of antibodies in the blood; in 1993, the Centers for Disease Control
defined AIDS to include all HIV-infected individuals with a CD4+ T-
lymphocyte count of less than 200 per microliter.
Case definition
• Whatever the definitions used in epidemiology, it is essential that
they be clearly stated, and easy to use and measure in a standard
manner in a wide variety of circumstances by different people.
• A clear and concise definition of what is considered a case ensures
that the same entity in different groups or different individuals is
being measured.
Measuring disease frequency
• Several measures of disease frequency are based on the concepts of
prevalence and incidence.
• An important factor in calculating measures of disease frequency is
the correct estimate of the numbers of people under study. Ideally
these numbers should only include people who are potentially
susceptible to the diseases being studied. For instance, men should
not be included when calculating the frequency of cervical cancer.
Population at risk
• The people who are susceptible to a given disease are called the
population at risk, and can be defined by demographic, geographic or
environmental factors.
• For instance, brucellosis occurs only among people handling infected
animals, so the population at risk consists of those working on farms
and in slaughterhouses (or hunters skinning the game).
Incidence and prevalence
• The incidence of disease represents the rate of occurrence of new
cases arising in a given period in a specified population, while
prevalence is the frequency of existing cases in a defined population at
a given point in time.
• These are fundamentally different ways of measuring occurrence and
the relation between incidence and prevalence varies among diseases.
• There may be low incidence and a high prevalence – as for diabetes –
or a high incidence and a low prevalence – as for the common cold.
• Colds occur more frequently than diabetes but last only a short time,
whereas diabetes is essentially lifelong.
Counting cases
• Measuring prevalence and incidence involves the counting of cases in defined
populations at risk. Reporting the number of cases without reference to the
population at risk can be used to give an impression of the overall magnitude of
a health problem, or of short-term trends in a population, for instance, during an
epidemic.
• WHO’s Weekly Epidemiological Record contains incidence data in the form of
case numbers, which in spite of their crude nature, can give useful information
about the development of epidemics of communicable diseases.
• The term “attack rate” is often used instead of incidence during a disease
outbreak in a narrowly-defined population over a short period of time. The
attack rate can be calculated as the number of people affected divided by the
number exposed, for example, in the case of a foodborne disease outbreak.
Rates
• Data on prevalence and incidence become much more useful if
converted into rates. A rate is calculated by dividing the number of
cases by the corresponding number of people in the population at risk
and is expressed as cases per 10n people.
• Some epidemiologists use the term “rate” only for measurements of
disease occurrence per time unit (week, year, etc.).
• Data on the population at risk are not always available and in many
studies the total population in the study area is used as an
approximation.
Prevalence
• Apart from age, several factors determine prevalence. In particular:
• • the severity of illness (if many people who develop a disease die
within a short time, its prevalence is decreased);
• • the duration of illness (if a disease lasts a short time its prevalence is
lower than if it lasts a long time);
• • the number of new cases (if many people develop a disease, its
prevalence is higher than if few people do so).
• P = Number of people with the disease or condition at a specified
time / Number of people in the population at risk at the specified time
(×10n )
Prevalence uses
• Since prevalence can be influenced by many factors unrelated to the
cause of the disease, prevalence studies do not usually provide strong
evidence of causality.
• Measures of prevalence are, however, helpful in assessing the need
for preventive action, healthcare and the planning of health services.
• Prevalence is a useful measure of the occurrence of conditions for
which the onset of disease may be gradual, such as maturity-onset
diabetes or rheumatoid arthritis.
Incidence
• Incidence refers to the rate at which new events occur in a population. Incidence
takes into account the variable time periods during which individuals are disease-
free and thus “at risk” of developing the disease.
• In the calculation of incidence, the numerator is the number of new events that
occur in a defined time period, and the denominator is the population at risk of
experiencing the event during this period. The most accurate way of calculating
incidence is to calculate what Last calls the “person-time incidence rate.” Each
person in the study population contributes one person-year to the denominator
for each year (or day, week, month) of observation before disease develops, or
the person is lost to follow-up.
• I = Number of new events in a specified period / Number of persons exposed to
risk during this period (×10n )
Example
• Since it may not be possible to measure disease-free periods precisely, the
denominator is often calculated approximately by multiplying the average size
of the study population by the length of the study period. This is reasonably
accurate if the size of the population is large and stable and incidence is low,
for example, for stroke.
• In a study in the United States of America, the incidence rate of stroke was
measured in 118 539 women who were 30–55 years of age and free from
coronary heart disease, stroke and cancer in 1976. A total of 274 stroke cases
were identified in eight years of follow-up (908 447 person-years). The overall
stroke incidence rate was 30.2 per 100 000 person-years of observation and
the rate was higher for smokers than non-smokers; the rate for ex-smokers
was intermediate.
Cumulative incidence
• Cumulative incidence is a simpler measure of the occurrence of a disease or
health status. Unlike incidence, it measures the denominator only at the
beginning of a study.
• Cumulative incidence is often presented as cases per 1000 population. The
cumulative incidence for stroke over the eight-year follow-up was 2.3 per 1000
(274 cases of stroke divided by the 118 539 women who entered the study).
• In a statistical sense, the cumulative incidence is the probability that
individuals in the population get the disease during the specified period.
• CI = Number of people who get a disease during a specified period / Number
of people free of the disease in the population at risk at the beginning of the
period (×10n )
Case fatality
• Case fatality is a measure of disease severity and is defined as the
proportion of cases with a specified disease or condition who die
within a specified time. It is usually expressed as a percentage.
• Case fatality (%) = Number of deaths from diagnosed cases in a
• given period / Number of diagnosed cases of the disease in the
• same period × 100
Interrelationships of the different measures
• Prevalence is dependent on both incidence and disease duration.
Provided that the prevalence (P) is low and does not vary significantly
with time, it can be calculated approximately as:
• P = incidence × average duration of disease
• The cumulative incidence rate of a disease depends on both the
incidence and the length of the period of measurement.
• Since incidence usually changes with age, age-specific incidence rates
need to be calculated. The cumulative incidence rate is a useful
approximation of incidence when the rate is low or when the study
period is short.
Comparing disease occurrence
• Measuring the occurrence of disease or other health states is the first step
of the epidemiological process.
• The next step is comparing occurrence in two or more groups of people
whose exposures have differed. An individual can be either exposed or
unexposed to a factor under study. An unexposed group is often used as a
reference group.
• The total amount of a factor that reaches an individual is called the “dose.”
• We can then compare occurrences to calculate the risk that a health effect
will result from an exposure. We can make both absolute and relative
comparisons; the measures describe the strength of an association between
exposure and outcome.
Absolute comparisons – risk difference
• The risk difference, also called excess risk, is the difference in rates of
occurrence between exposed and unexposed groups in the
population.
• It is a useful measure of the extent of the public health problem
caused by the exposure.
• For example, the risk difference between the incidence rate of stroke
in women who smoke, and the rate of stroke in women who have
never smoked, is 31.9 per 100 000 person-years.
Absolute comparisons – Attributable fraction
(exposed)
• The attributable fraction (exposed), also known as the etiological
fraction (exposed), is the proportion of all cases that can be attributed
to a particular exposure.
• We can determine the attributable fraction (AF) by dividing the risk
(or attributable) difference by the incidence among the exposed
population.
• The attributable fraction of smoking for stroke in the smokers is:
((49.6 – 17.7)/49.6) ×100 = 64%.
• What are the practical implications?
Practical considerations
• When a particular exposure is believed to be a cause of a given disease, the
attributable fraction is the proportion of the disease in the specific population that
would be eliminated if the exposure were eliminated. In the above example, one
would expect to achieve a 64% reduction in the risk of stroke among the women
smokers if smoking were stopped, based on the assumption that smoking is both
causal and preventable.
• Attributable fractions are useful for assessing priorities for public health action.
• For example, both smoking and air pollution are causes of lung cancer, but the
attributable fraction due to smoking is usually much greater than that due to air
pollution. Only in communities with very low smoking prevalence and severe air
pollution is the latter likely to be the major cause of lung cancer. In most countries,
smoking control should take priority in lung cancer prevention programmes.
Absolute comparisons – population
attributable risk
• The population attributable risk (PAR) is the incidence of a disease in a
population that is associated with (or attributed to) an exposure to a
risk factor. This measure is useful for determining the relative
importance of exposures for the entire population.
• It is the proportion by which the incidence rate of the outcome in the
entire population would be reduced if exposure were eliminated.
• PAR can be estimated by the formula: PAR = Ip − Iu / Ip,
• where Ip is the incidence of the disease in the total population and Iu
is the incidence of the disease among the unexposed group.
Relative comparisons – relative risk
• The relative risk (also called the risk ratio) is the ratio of the risk of
occurrence of a disease among exposed people to that among the
unexposed.
• The risk ratio of stroke in women who smoke, compared with those who
have never smoked, is 2.8 (49.6 /17.7).
• The risk ratio is a better indicator of the strength of an association than the
risk difference, because it is expressed relative to a baseline level of
occurrence.
• Unlike the risk difference, it is related to the magnitude of the baseline
incidence rate; populations with similar risk differences can have greatly
differing risk ratios, depending on the magnitude of the baseline rates.
Practical implications
• The risk ratio is used in assessing the likelihood that an association
represents a causal relationship.
• For example, the risk ratio of lung cancer in long-term heavy smokers
compared with non-smokers is approximately 20. This is very high and
indicates that this relationship is not likely to be a chance finding. Of
course, smaller risk ratios can also indicate a causal relationship, but
care must be taken to eliminate other possible explanations.
Attributable risk
• Attributable risk is the rate (proportion) of a disease or other outcome in
exposed individuals that can be attributed to the exposure.
• This is a more useful term for public health purposes as it reflects the
amount, usually expressed as a percentage, by which the risk of a disease
is reduced by elimination or control of a particular exposure. Using
attributable risk, it is possible to estimate the number of people spared
the consequences of exposure, by subtracting the rate of the outcome
(usually incidence or mortality) among the unexposed from the rate
among the exposed individuals.
• For example, if there were 6 deaths per 100 among smokers, and 1 death
per 100 in non-smokers, the attributable risk would be 5 per 100.